[IPython-dev] nbconvert on Scientific Linux 5

[Jon Wilson]

Hi Aaron,
I solved that problem.  The issue was that cipher-aes uses the AESNI 
instruction set, but the CPUs in the machine I'm on predate that 
instruction set.  Luckily there was a C flag that I could unset in order 
to force the use of a software-only implementation.

The question of AES crypto as a dependency of pandoc is symptomatic of 
what is my growing opinion of Haskell packages: insufficient 
modularity.  Haskell-platform won't build unmodified without OpenGL, yet 
there are many things one might want to do with Haskell that work fine 
without OpenGL.  Similarly, pandoc doesn't use crypto, but its build 
fails if unrelated packages won't build.  Quite frustrating.

Now my problem is with the builtin regex library, which complains of 
undefined symbols at the linker stage.  I thought that it couldn't find 
the PCRE C library, but that is apparently not the case.  I found an 
issue on github, so we'll see if any solutions are forthcoming from the 
developer.
Regards,
Jon

On 06/28/2014 07:48 PM, Aaron Meurer wrote:
> What was the error message?
>
> As I mentioned, Pandoc has a ton of recursive dependencies. cipher-aes
> is probably a dependency of a dependency.
>
> Aaron Meurer